Introducing GPT-4.5 "Orion"

OpenAI unveils GPT-4.5, codenamed "Orion," its largest AI language model to date, enhancing user interactions.

Enhanced User Interaction

GPT-4.5 offers improved writing, expanded knowledge, and a refined conversational tone for natural, emotionally nuanced responses.

Availability and Access

Initially available as a research preview, GPT-4.5 is accessible to ChatGPT Pro subscribers via a $200 monthly plan.​

Performance Highlights

While excelling in language tasks, GPT-4.5 underperforms in math and science compared to smaller models like o3-mini.

Training and Development

Trained using advanced supervision techniques, GPT-4.5 combines traditional methods with new approaches for better accuracy.​

Cost and Resource Implications

Operating GPT-4.5 is costly, with expenses reaching $75 per million input tokens and $150 per million output tokens.

Competitive Landscape

Despite advancements, GPT-4.5 faces competition from models offering comparable or superior performance at lower costs.

Future Developments

OpenAI plans to launch GPT-5 by May, integrating the new o3 reasoning model to enhance performance and capabilities.

Not a Frontier AI Model

OpenAI emphasizes that GPT-4.5 is not classified as a frontier AI model, acknowledging certain limitations.

Improved Accuracy and Interaction

GPT-4.5 exhibits fewer hallucinations and provides warmer, more emotionally nuanced interactions than previous versions.